samtools flagstat output

[manore]

Hi,
I mapp pair-end reads to a reference genome (stampy) then I convert SAM file into BAM and finally I get some stats using the samtools flagstat.
I get a file with a content similar to:
4198456 + 0 in total (QC-passed reads + QC-failed reads)
0 + 0 duplicates
4022089 + 0 mapped (95.80%:-nan%)
4198456 + 0 paired in sequencing
2099228 + 0 read1
2099228 + 0 read2
3796446 + 0 properly paired (90.42%:-nan%)
4013692 + 0 with itself and mate mapped
8397 + 0 singletons (0.20%:-nan%)
167574 + 0 with mate mapped to a different chr
72008 + 0 with mate mapped to a different chr (mapQ>=5)

I don't understand the detailed explanation of samtools flagstat output.
* What does "singleton" mean ? This means that one reads of the paired end reads is mapped to the genome
* What "with itself and mate mapped" mean? It means that both reads mapped to genome

[aituka]

Hi did you find something?
i have the same questions here.
thanks
tuka

[twaddlac]

Both of your answers are correct.This forum might shed some light on other questions about the output that you may have.

http://ppotato.wordpress.com/2010/08/25/samtool-bitwise-flag-paired-reads/

I hope this helps!

[vkartha]

samtools flagstat dead?

Both of your answers are correct.This forum might shed some light on other questions about the output that you may have.

http://ppotato.wordpress.com/2010/08/25/samtool-bitwise-flag-paired-reads/

I hope this helps!

I look now in the samtools manual and there is no longer a flagstat option to use for estimating these stats. Can anyone tell me why? I used it earlier to see how 'well' paired-end reads aligned to the genome looking at the 'properly-paired %' statistic which was a part of the samtools flagstat output.

Does anyone know why? Thanks

[Richard Finney]

The manual (at http://samtools.sourceforge.net/samtools.shtml ) does not document flagstat, but it is realy there ...

-bash-3.00$ ~/samtools-0.1.18/samtools flagstat
Usage: samtools flagstat <in.bam>

The "NEWS" files says flagstat was added in ver 0.1.3 on 15 April, 2009

[vkartha]

assessing how -r affects Tophat output

The manual (at http://samtools.sourceforge.net/samtools.shtml ) does not document flagstat, but it is realy there ...

-bash-3.00$ ~/samtools-0.1.18/samtools flagstat
Usage: samtools flagstat <in.bam>

The "NEWS" files says flagstat was added in ver 0.1.3 on 15 April, 2009

Thanks for that Richard! Maybe you will be able to help me with my problem!

So to briefly outline what I did - I wanted to see how using different -r options for Tophat2 will affect my alignment (part of an RNASeq study).

I initially took a subset (5 million) 101 bp long paired-end reads from 4 control and 4 disease samples and mapped them to the ref transcriptome using Bowtie2.

On doing so, I then used picard tools on the output sam files to first sort them and then estimate the insert size statistics. This gave me the mean and standard deviation fragment length based on the alignment so I had to subtract twice the read length to get Tophat's 'inner distance' between pairs option value.

So what for this analysis was I took the average of the 4 control means, the average of the 4 disease means (-15 and -33 respectively) and a high and low extreme value (-50 and +50) just to see how it would affect my alignment. I chose a common std deviation of 55 and aligned 1 chosen disease sample to the ref genome using Tophat2 and these 4 different -r values, each a single run.

Coming back to this thread's topic, I then used samtools flagstat to evaluate how well the -r option worked for the alignment looking at the 'properly-paired %' stat which is part of the output (I read that this is a common procedure I'm not 100% sure if it's valid).

Quite to my surprise, the mean disease -r I mentioned earlier (-33) gave a % of only 82 while the high extreme value of +50 gave the highest % of 92. Why is this??? +50 is nowhere close to the mean I had estimated using picard tools.

Please do help and I really appreciate your prompt response thus far.

[Richard Finney]

My guess ...
The larger expected mate inner distance (tophat -r parameter) allows it to look farther out in order to align the weaker of the two pairs. The result is more alignments.

[vkartha]

My guess ...
The larger expected mate inner distance (tophat -r parameter) allows it to look farther out in order to align the weaker of the two pairs. The result is more alignments.

What do you mean by "weaker of the two pairs"? And if that is the case - how would I estimate what the right -r value to use would be, given it's so far off from what the actual estimated mean is? Using that with Tophat just doesn't make sense when it expects the 'mean' inner distance between mate pairs

[Richard Finney]

I haven't used tophat for several years though was impressed when it first came out.
It might be doing a strategy of if one pair has a perfect match, look nearby with a range for the to place the other "weaker" pair which is not perfect matched. If the range is bigger (i.e. the expected distance is bigger), it is more likely to place the second pair. This is speculation, I don't know what strategy it uses. This is a bigger deal with rna-seq and gene exon models where you'll have exon skipping.

[paulorapazote]

samtools flagstat output explanation

Hi

I am trying to find a detailed explanation to the samtools flagstat output, without success. Even in http://samtools.sourceforge.net/ there is no mention to the flagstat command...

Any help?

Thanks in advance.

Paulo

[westerman]

I think that the third Google hit I get is rather good.

http://www.biostars.org/p/12475/

If you have a specific question about flagstat then ask it. I agree that the samtools doc itself should talk more about flagstat. Perhaps the developers thought that the output was too obvious to mention?

[sdriscoll]

If you have gone through the trouble of writing code to produce properly formatted SAM output for paired end alignments then, and only then, is the flagstat output obvious.